Noncausal Predictive Image Coding Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image and video coding technologies face challenges in achieving fully noncausal predictive coding due to interdependence issues, which result in high computational costs and limited accuracy, especially with large matrices involved in two-sided noncausal residual signal processing.
Innovation Solution
The method involves splitting images or video into blocks, extending them, and using selected intra-block or path noncausal predictors for encoding and decoding, employing techniques like direct spatial, matrix-vector, DFT, and symmetric convolution encoding and decoding to generate and reconstruct two-sided residual signals with reduced computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fully noncausal predictive coding is implemented using two-sided residual signal processing, then coding accuracy is improved, but computational cost increases due to large matrices
Solution Approach 1:
The patent divides the image or video signal into multiple blocks or paths, processing each block independently with its own smaller matrix operations. This segmentation approach maintains the benefits of noncausal prediction while reducing the computational burden of large matrix operations by breaking them into smaller, manageable segments that can be processed in parallel or sequentially with lower memory requirements.
2Measurement precision
If large matrices are used for two-sided noncausal residual signal processing, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
By segmenting the processing into smaller blocks with individual prediction matrices, the patent reduces device complexity while maintaining overall prediction accuracy. Each block uses a smaller, simpler matrix operation rather than one large complex matrix operation across the entire image or frame.
3Device complexity
If blocks are processed independently, then computational complexity is reduced, but interdependence information is lost
Solution Approach 1:
The patent performs preliminary actions by extending blocks with boundary pixels before independent processing, ensuring that each block has access to relevant neighboring information. This preliminary preparation allows independent block processing to maintain accuracy by pre-including boundary information that would otherwise require complex inter-block dependencies.
Data Source
AI summary
The present invention presents fully noncausal predictive encoding and decoding methods for image, video and other signal coding. The presented noncausal predictive image coding methods largely reduce the prohibitive computational cost of the prior invention. The presented noncausal signal encoding method comprises: (1) splitting the source signal into a plurality of noncausal coding units; (2) extending each noncausal coding unit with the selected extension type; and (3) encoding each noncausal coding unit with the selected intra-unit noncausal predictor and intra-unit noncausal predictive encoding method.